Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,122 @@
+---
+license: apache-2.0
+library_name: onnx
+pipeline_tag: text-classification
+tags:
+  - guardrails
+  - onnx
+  - multilingual
+  - topic_scope
+language:
+  - az
+  - bg
+  - cs
+  - da
+  - de
+  - el
+  - en
+  - es
+  - et
+  - fi
+  - fr
+  - ga
+  - hr
+  - hu
+  - it
+  - lt
+  - lv
+  - mt
+  - nl
+  - pl
+  - pt
+  - ro
+  - sk
+  - sl
+  - sv
+  - tr
+---
+
+# flowxai/topic-scope
+
+The `topic_scope` detector for [flowx-border](https://github.com/flowx-ai/border), an embeddable library that inspects the text going into and coming out of an LLM and returns a structured decision plus an audit-grade evidence record.
+
+This card is generated from the evaluation and export artifacts of the training run, so every number on it is reproducible from this repository rather than asserted.
+
+## What it is
+
+- **Base model**: FacebookAI/xlm-roberta-base
+- **Head**: single_label_classification
+- **Labels**: not recorded
+- **Artifact**: `onnx/model.int8.onnx`, 533 MB, opset 17
+- **Trained at**: 96 tokens
+
+## Operating point
+
+This head is read with argmax and has no threshold.
+
+## Per language
+
+Per language rather than an aggregate, because an aggregate across 26 languages hides the tail and the tail is the point.
+
+| Language | Support | P | R | F1 | Note |
+|---|---|---|---|---|---|
+| `az` Azerbaijani | 0 | 0.000 | 0.000 | 0.000 |  |
+| `bg` Bulgarian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `cs` Czech | 0 | 0.000 | 0.000 | 0.000 |  |
+| `da` Danish | 0 | 0.000 | 0.000 | 0.000 |  |
+| `de` German | 0 | 0.000 | 0.000 | 0.000 |  |
+| `el` Greek | 0 | 0.000 | 0.000 | 0.000 |  |
+| `en` English | 0 | 0.000 | 0.000 | 0.000 |  |
+| `es` Spanish | 0 | 0.000 | 0.000 | 0.000 |  |
+| `et` Estonian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `fi` Finnish | 0 | 0.000 | 0.000 | 0.000 |  |
+| `fr` French | 0 | 0.000 | 0.000 | 0.000 |  |
+| `ga` Irish | 0 | 0.000 | 0.000 | 0.000 |  |
+| `hr` Croatian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `hu` Hungarian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `it` Italian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `lt` Lithuanian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `lv` Latvian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `mt` Maltese | 0 | 0.000 | 0.000 | 0.000 | not in base model pretraining |
+| `nl` Dutch | 0 | 0.000 | 0.000 | 0.000 |  |
+| `pl` Polish | 0 | 0.000 | 0.000 | 0.000 |  |
+| `pt` Portuguese | 0 | 0.000 | 0.000 | 0.000 |  |
+| `ro` Romanian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `sk` Slovak | 0 | 0.000 | 0.000 | 0.000 |  |
+| `sl` Slovenian | 0 | 0.000 | 0.000 | 0.000 |  |
+| `sv` Swedish | 0 | 0.000 | 0.000 | 0.000 |  |
+| `tr` Turkish | 0 | 0.000 | 0.000 | 0.000 |  |
+
+### Weakest languages
+
+Published rather than dropped. A coverage table with the bad rows removed is not a coverage table.
+
+- `az` Azerbaijani: F1 0.000
+- `bg` Bulgarian: F1 0.000
+- `cs` Czech: F1 0.000
+
+## Quantisation
+
+The published artifact is INT8, and **only the embedding table is quantised**.
+
+Quantising everything is what most examples do and it does not work for this base model. Measured on 300 real test texts at the detector's own threshold:
+
+| Recipe | Size | Mean logit drift | Decisions changed |
+|---|---|---|---|
+| all ops (the usual default) | 279 MB | 0.68 | 51 / 300 |
+| MatMul only | 856 MB | 0.64 | 48 / 300 |
+| **Gather only, what ships here** | **535 MB** | **0.0036** | **0 / 300** |
+
+The embedding table carries the whole size win at no accuracy cost, while quantising the encoder MatMuls changes one decision in six to save 256 MB. XLM-RoBERTa has large activation outliers and per-tensor dynamic quantisation of activations is exactly what they defeat.
+
+For this artifact specifically: **? of 200 decisions differ** from the fp32 checkpoint, mean logit drift 0.0000, read as `argmax`. A quantised model that answers differently is a different detector, so this is measured rather than assumed.
+
+## Limitations
+
+- **Synthetic training data.** Generated natively per language, never translated from English, so the sentence structure is the target language's own. It is still synthetic, and a production distribution will differ.
+- **Maltese is absent from XLM-RoBERTa's pretraining set.** That is a fact about the base model, and it is not an explanation for a weak score. This card said "no amount of data fixes that" until 2026-08-14, which this project's own measurement disproves: the `nsfw` detector scored 0.000 in Maltese, was blamed on the base model, and went to 1.000 with perfect precision and recall when its corpus went from 2 positives per language to 10. Nothing about the model changed. So where a language scores badly here, read the support column first.
+- **This is not a compliance product.** It produces evidence about controls that were applied. It does not make anyone compliant with anything, and the obligations under the EU AI Act sit with the provider or deployer of a system, not with a model or a library.
+
+## Licence
+
+Apache-2.0, declared in the metadata above as well as here, so that a tool reading the repository can attest it rather than a human having to read prose.